I'm using NLOpt 2.4.2 and noticed that the output is only the last values and not the optimal values. The problem is setup using the C++ API with inequality constraints and the COBYLA algorithm. When the x values, cost, and whether the constraints pass are printed in the objective callback and constraint callback; neither the less optimal constraint passing solution nor the more optimal constraint failing solution is returned. Instead, it returns a constraint failing suboptimal solution, which just happens to be the last solution. Has anyone else seen this behavior?
I'm using NLOpt 2.4.2 and noticed that the output is only the last values and not the optimal values. The problem is setup using the C++ API with inequality constraints and the COBYLA algorithm. When the x values, cost, and whether the constraints pass are printed in the objective callback and constraint callback; neither the less optimal constraint passing solution nor the more optimal constraint failing solution is returned. Instead, it returns a constraint failing suboptimal solution, which just happens to be the last solution. Has anyone else seen this behavior?